Histogram analysis of diffusion kurtosis magnetic resonance imaging in differentiation of pathologic Gleason grade of prostate cancer

• Published in: Urologic oncology 2015-08, Vol.33 (8), p.337.e15-337.e24
• Main Authors: Wang, Qing, M.M, Li, Hai, M.M, Yan, Xu, Ph.D, Wu, Chen-Jiang, M.M, Liu, Xi-Sheng, M.D, Shi, Hai-Bin, M.D, Zhang, Yu-Dong, M.D
• Format: Article
• Language: English
• Subjects: Aged; Aged, 80 and over; Cell Differentiation; Dapp; Diffusion Magnetic Resonance Imaging - methods; DKI; Histogram analysis; Humans; Kapp; Male; Middle Aged; Neoplasm Grading; Prostate cancer; Prostatic Neoplasms - diagnostic imaging; Prostatic Neoplasms - pathology; Radiography; Urology
• ISSN: 1078-1439; 1873-2496
• DOI: 10.1016/j.urolonc.2015.05.005

Abstract Objective To investigate diagnostic performance of diffusion kurtosis imaging with histogram analysis for stratifying pathologic Gleason grade of prostate cancer (PCa). Materials and methods This retrospective study was approved by the institutional review board, and written informed consent was waived. A total of 110 patients pathologically confirmed as having PCa (diameter>0.5 cm) underwent preoperative diffusion-weighted magnetic resonance imaging ( b value of 0–2,100 s/mm2 ) at 3 T. Data were postprocessed by monoexponential and diffusion kurtosis models for quantitation of apparent diffusion coefficients (ADCs), apparent diffusion for Gaussian distribution ( Dapp ), and apparent kurtosis coefficient ( Kapp ). The measurement was based on an entire-tumor histogram analysis approach. The ability of imaging indices for differentiating low-grade (LG) PCa (Gleason score [GS]≤6) from intermediate-/high-grade (HG: GS>6) PCa was analyzed by receiver operating characteristic regression. Results There were 49 LG tumors and 77 HG tumors at pathologic findings. HG-PCa had significantly lower ADCs, lower diffusion kurtosis diffusivity ( Dapp ), and higher kurtosis coefficient ( Kapp ) in mean, median, 10th, and 90th percentile, with higher Dapp in skewness and kurtosis than LG-PCa ( P < 0.05). The 10th ADCs, the 10th Dapp , and the 90th Kapp showed relatively higher area under receiver operating characteristic curve (Az), Youden index, and positive likelihood ratio in stratifying aggressiveness of PCa against other indices. The 90th Kapp showed relatively higher correlation ( ρ >0.6) with ordinal GS of PCa; significantly higher Az, sensitivity, and specificity (0.889, 74.1%, and 93.9%, respectively) than the 10th Dapp did (0.765, 61.0%, and 79.6%, respectively; P